spinf-serve 0.1.1__tar.gz

spinf_serve-0.1.1/.github/workflows/ci.yml

@@ -0,0 +1,29 @@
+# The CPU suite and lint on every push and pull request. The GPU parity gate (README "Develop") runs by hand on a card: there is
+# no GPU runner. The official packs' golden fixtures need a checkout of spinf-prompts next to the tests (SPINF_PROMPTS_DIR);
+# without one those tests skip. SPINF_PROMPTS_REPO (a repository variable) and SPINF_PROMPTS_TOKEN (a secret) point a fork or a
+# mirror at its own copy of the pack repository; unset, the public one is used.
+name: ci
+on:
+  push:
+  pull_request:
+  workflow_dispatch:
+permissions:
+  contents: read
+jobs:
+  test:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v7
+      - uses: actions/checkout@v7
+        id: prompts
+        continue-on-error: true
+        with:
+          repository: ${{ vars.SPINF_PROMPTS_REPO || 'spinfinc/spinf-prompts' }}
+          token: ${{ secrets.SPINF_PROMPTS_TOKEN || github.token }}
+          path: .spinf-prompts
+      - uses: astral-sh/setup-uv@v10.2.0
+      - run: uv sync --frozen --extra server
+      - run: uv run ruff check . && uv run ruff format --check .
+      - run: uv run pytest
+        env:
+          SPINF_PROMPTS_DIR: ${{ steps.prompts.outcome == 'success' && format('{0}/.spinf-prompts', github.workspace) || '' }}

spinf_serve-0.1.1/.gitignore

@@ -0,0 +1,10 @@
+.venv/
+__pycache__/
+*.pyc
+.pytest_cache/
+.ruff_cache/
+dist/
+.env
+# the maintainers' news fixtures carry article text that is not cleared for publication: local only (the public fixture is
+# fixtures/wikinews_*.json.gz)
+fixtures/news_*

spinf_serve-0.1.1/.python-version

@@ -0,0 +1 @@
+3.12

spinf_serve-0.1.1/ACKNOWLEDGEMENTS.md

@@ -0,0 +1,57 @@
+# Acknowledgements
+
+spinf-serve stands on other people's open work. This file names it: the models we run and re-quantize, the projects we
+build on, the techniques we implement, and the data we checked our calibrations and prompts on. The legal attributions
+for code included in this repository are in `NOTICE`; each checkpoint we publish credits its source in its model card
+(`quantizations/*/MODEL_CARD.md`). The prompt packs have their own acknowledgements in the spinf-prompts repository.
+
+Rule for contributors: the change that adds a dependency, vendors code, adopts a recipe, adds a model or checks results
+on a dataset also adds the line here (and in `NOTICE` when code is included).
+
+## Models and checkpoints
+
+| what | by | terms | how we use it |
+|---|---|---|---|
+| [Gemma 4](https://huggingface.co/google) 12B and 31B | Google DeepMind | Apache 2.0 (Gemma 4; Google's licence page: ai.google.dev/gemma/docs/gemma_4_license) | the `spinf/gemma-4-12b-fp8` / `spinf/gemma-4-31b-fp8` models: quantized copies with the original notices retained |
+| [Qwen3.8-27B](https://huggingface.co/Qwen/Qwen3.8-27B) | Alibaba Qwen team | Apache 2.0 | the `spinf/qwen3.8-27b-int4` model (our GPTQ int4 copy) |
+| [pplx-decider-v1-27b](https://huggingface.co/perplexity-ai/pplx-decider-v1-27b) | Perplexity AI | Apache 2.0 (derived from Qwen3.8-27B, revision 1d4bf0f2) | the decision model with a 255-code readout; our GPTQ int4 copy |
+| [btbtyler09/Qwen3.8-27B-GPTQ-4bit](https://huggingface.co/btbtyler09/Qwen3.8-27B-GPTQ-4bit) | btbtyler09 | see the repo | the GPTQ recipe we adopted (GPTQModel 6.0.3, group 32, FOEM) and the checkpoint we served before publishing our own |
+
+## Projects
+
+- [PyTorch](https://pytorch.org) (inductor, FlexAttention, `torch.compile`), [Triton](https://github.com/triton-lang/triton): the kernels.
+- [Hugging Face](https://huggingface.co) transformers, tokenizers, safetensors, huggingface_hub: model definitions (the Gemma 4 and
+  Qwen3.5 modules, the vision towers), tokenization, weights.
+- [flash-linear-attention](https://github.com/fla-org/flash-linear-attention) (Songlin Yang, Yu Zhang and contributors): the Gated
+  DeltaNet chunk kernels; one state kernel is vendored with changes (`NOTICE`).
+- [GPTQModel](https://github.com/modelcloud/gptqmodel) (ModelCloud): the int4 quantization of our Qwen-architecture checkpoints.
+- [vLLM](https://github.com/vllm-project/vllm): the reference for the fp8 scheme's numerics (`--quantization fp8`), the engine we
+  compare throughput against, and the OpenAI-compatible server shape we follow.
+- [FastAPI](https://fastapi.tiangolo.com), [uvicorn](https://www.uvicorn.org), [pydantic](https://docs.pydantic.dev): the server and the protocol models.
+
+## Techniques
+
+- GPTQ (Frantar, Ashkboos, Hoefler, Alistarh, 2022) and FOEM (First-Order Error Matters, AAAI 2026): the int4 weights.
+- LLM.int8() (Dettmers, Lewis, Belkada, Zettlemoyer, 2022): the mixed-precision decomposition our i8 split scheme fixes in advance
+  (the outlier channels kept exact, the rest int8 per token).
+- SmoothQuant (Xiao et al., 2022): evaluated as an alternative to the outlier split (docs/findings/QWEN.md).
+- Gated DeltaNet (Yang, Kautz, Hatamizadeh, 2024): the recurrent layers of the Qwen3.5 architecture.
+- FlexAttention (PyTorch): the tree masks that let many prompts share one read of the content.
+
+## Data
+
+Used to calibrate the fp8 activation scales and the outlier channels and to check accuracy; never bundled, never used to
+train anything that ships in this repository.
+
+| dataset | used for | terms |
+|---|---|---|
+| [Wikinews](https://en.wikinews.org) articles | the public parity fixture and benchmark job set (`fixtures/`, 24 articles; each one's title, URL and revision in the fixture's `sources`) | CC BY 2.5 |
+| Imagenette and Imagewoof (fast.ai) | image calibration and accuracy suites | Apache 2.0 |
+| UCF101 | video suites | see the dataset's terms |
+| ESC-50 | audio calibration and suite | CC BY-NC 3.0 (calibration and checks only) |
+| GTZAN | audio suite | see the dataset's terms |
+| RAGTruth | the outlier-channel calibration stream; a decision benchmark | see the dataset's terms |
+| evol-codealpaca-v1 and C4 | GPTQ calibration samples (the adopted recipe) | see each dataset's terms (C4: ODC-BY) |
+| the decision benchmarks of the pplx-decider card (Belebele, TruthfulQA, WinoGrande, Circa, FinancialPhraseBank, TabFact, RAGTruth, ContractNLI, JudgeBench, BBH, JevBench) | accuracy checks of the Qwen-architecture models | each set's own terms |
+
+The news and moderation texts of our internal test bed are not redistributed; their sources are listed in that repository.

spinf_serve-0.1.1/LICENSE

@@ -0,0 +1,202 @@
+
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright 2026 spinf Inc.
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

spinf_serve-0.1.1/NOTICE

@@ -0,0 +1,11 @@
+spinf-serve
+Copyright 2026 spinf Inc.
+
+This product includes software developed by third parties:
+
+- flash-linear-attention (fla-core 0.5.2), MIT License, Copyright (c) 2023-2026 Songlin Yang, Yu Zhang, Zhiyuan Li.
+  spinf_serve/models/qwen35/gdn.py vendors its Gated DeltaNet chunk state kernel (chunk_gated_delta_rule_fwd_kernel_h_blockdim64)
+  with changes (initial and final states addressed by row index).
+
+The model checkpoints this engine runs are distributed by their authors under their own terms (see ACKNOWLEDGEMENTS.md);
+none are included in this repository.

spinf_serve-0.1.1/PKG-INFO

@@ -0,0 +1,240 @@
+Metadata-Version: 2.5
+Name: spinf-serve
+Version: 0.1.1
+Summary: The open-source inference engine for decision models: one input read once, hundreds of questions scored in parallel on one GPU
+License-Expression: Apache-2.0
+License-File: LICENSE
+License-File: NOTICE
+Requires-Python: >=3.12
+Requires-Dist: pydantic<3,>=2.11
+Provides-Extra: engine
+Requires-Dist: flash-linear-attention==0.5.2; extra == 'engine'
+Requires-Dist: huggingface-hub==1.30.0; extra == 'engine'
+Requires-Dist: numpy==2.5.2; extra == 'engine'
+Requires-Dist: pillow>=11; extra == 'engine'
+Requires-Dist: safetensors==0.8.0; extra == 'engine'
+Requires-Dist: tokenizers==0.23.2; extra == 'engine'
+Requires-Dist: torch==2.14.0; extra == 'engine'
+Requires-Dist: transformers==5.16.1; extra == 'engine'
+Requires-Dist: triton==3.8.0; extra == 'engine'
+Provides-Extra: server
+Requires-Dist: fastapi>=0.118; extra == 'server'
+Requires-Dist: flash-linear-attention==0.5.2; extra == 'server'
+Requires-Dist: huggingface-hub==1.30.0; extra == 'server'
+Requires-Dist: numpy==2.5.2; extra == 'server'
+Requires-Dist: pillow>=11; extra == 'server'
+Requires-Dist: safetensors==0.8.0; extra == 'server'
+Requires-Dist: tokenizers==0.23.2; extra == 'server'
+Requires-Dist: torch==2.14.0; extra == 'server'
+Requires-Dist: transformers==5.16.1; extra == 'server'
+Requires-Dist: triton==3.8.0; extra == 'server'
+Requires-Dist: uvicorn>=0.37; extra == 'server'
+Description-Content-Type: text/markdown
+
+# spinf-serve
+
+**The open-source inference engine for decision models.**
+
+spinf-serve reads a piece of content once and answers hundreds of questions about it in parallel, as probabilities
+rather than text. It turns a standard open model (Gemma 4, Qwen3.8) into a decision model through **decision mode**:
+one input, many questions, scores instead of generation, and the same score whatever else is in the batch. It runs on
+one 24 GB GPU and serves an OpenAI-compatible API that is also the API of the hosted spinf service.
+
+## What it does
+
+- **Scoring.** A question is a template that ends where the answer goes, with the answers to score there. The engine
+  returns the probability of each answer, its log-probability, and the same question's probability on empty content
+  (the floor), so a score can be read as a lift over the question's own bias.
+- **Many questions, one read.** The content is read once; every question is a branch of the same pass. Adding a
+  question costs its own tokens only. Placeholders turn one template into a grid of questions.
+- **Extraction and generation** from the same read: `{*}` text, `{n}` numbers and `{$}` amounts decoded with their
+  probabilities, or free text generated into a slot.
+- **Batch invariance.** With the calibrated fp8 scheme, an input's scores are bit-identical whether it is scored
+  alone or packed with thousands of others. Scores compare across calls, days and servers; a `system_fingerprint`
+  says when they do not.
+- **Media.** Images, audio and video frames in the content, read by the models that support them.
+- **Prompt packs.** Ready-made, calibrated question sets preloaded into the server and named in a request
+  ([below](#prompt-packs)).
+- **Throughput.** Inputs are packed into 16k-token passes on one GPU: about 10,000 tokens per second on an RTX 4090
+  with the 12B model.
+- **One API, hosted or self-hosted.** The request and response are those of `api.spinf.com`; a client switches by
+  changing the base URL and the key.
+
+## Speed and scale
+
+spinf-serve exists for one workload: **many questions about one content**. Our own first use, a daily oil-market
+index read from 15 million news articles, taught us that asking thousands of narrow, calibrated questions of every
+article gives a far more accurate read than any single prompt, and that no engine built for chat could do it at a
+price that made sense. A chat engine re-reads the content for every question; even with prefix caching it generates
+one token at a time per prompt and treats every question as a separate request. spinf-serve reads the content once
+and lays every question out as a branch of that read, packed with other inputs into 16k-token passes on one GPU.
+
+Measured on one RTX 4090 with Gemma 4 12B in fp8, the same 1,344 questions (24 news articles of about 470 tokens, 56
+questions each, every answer a probability over two options: the public job set in `fixtures/`) sent over HTTP to
+spinf-serve and to **vLLM 0.26 with prefix caching**, which is what most people would reach for (`benchmarks/`):
+
+| one RTX 4090, warm, per second | spinf-serve | vLLM 0.26, prefix caching, 128 requests in flight |
+|---|---|---|
+| questions answered | **491** | 200 |
+| tokens of work processed (content once, then each question's own tokens) | 11,200 | 4,600 |
+| the same result for an engine that re-reads the content per question | 236,000 tokens | 96,000 tokens |
+| the 1,344 questions | **2.7 s** | 6.7 s |
+
+2.5 times faster on the same card for the same answers, and vLLM is already doing the right thing here: its prefix
+cache reuses each article across its 56 questions. What it cannot do is answer the questions in one forward pass, so it
+runs 1,344 single-token requests against the cache, while spinf-serve lays the 56 questions out as branches of one read
+and packs the 24 articles into two 16k-token passes. The gap grows with the number of questions per content: at 56
+questions the shared read saves 21 times the tokens; at a thousand it saves hundreds of times, and a thousand questions
+on one content still cost about what the questions themselves weigh. Processed-token throughput stays at the engine's
+rate, about 11,000 per second per 4090 for the 12B model (on an earlier set of 18 shorter articles with 74 questions each
+the ratio was 2.9). The int4 Qwen3.8 27B model showed the same ratio range (2.2 to 2.8 times vLLM with prefix caching) on
+2,211 decision-benchmark jobs in our internal measurements. Anyone can rerun the comparison: `benchmarks/README.md`.
+
+What this means for a job:
+
+- **Content moderation**: the `spinf/moderation` pack asks 106 questions of every message (22 categories, three
+  decision modes). One 4090 moderates about 20,000 messages per hour through the full pack, each message scored for
+  every category in one read.
+- **News and document analysis**: 74 calibrated questions per article, aggregated into indexes, at nearly 700
+  questions per second per GPU.
+- **Triage, routing, tagging**: intent in two levels, urgency, sentiment, risk flags, personal-data checks and a
+  policy check in one pass, each a question in the same list, each costing its own tokens only.
+
+**Extraction and generation** ride the same read but add decoding steps: a searched slot (`{*}` `{n}` `{$}`) costs
+about as much GPU time as 200 to 600 tokens of scoring, a generated slot up to about 2,000. They still benefit from
+the shared read and the packing: the content and the questions are not re-read, the slots of many inputs are decoded
+together, and a few slots next to hundreds of scored questions change the pass's cost only a little.
+
+The example packs (triage, product listings, entertainment metadata, reviews, brand analysis) are in
+[spinf-prompts](https://github.com/spinfinc/spinf-prompts).
+
+## Prompt packs
+
+A specialised use case is rarely one question: content moderation is 106 of them, each calibrated against normal
+traffic and combined into category scores and a decision. spinf-serve standardises a format for this: a **prompt pack**
+is one file holding a set of questions, how to present the content to the model, how the answers turn into scores,
+and how the scores turn into a decision. Packs are preloaded into the server, so a request names one (`"packs": [{"id":
+"spinf/moderation"}]`) instead of carrying hundreds of questions over and over, and gets the pack's categories and
+decision back next to its own questions, from the same read of the content.
+
+Because a pack is a file, it is also a **ready-to-go, shareable set**: the server loads packs from
+[spinf-prompts](https://github.com/spinfinc/spinf-prompts), the official repository (content moderation first), and
+from any GitHub repository or local folder you point it at, enabling the ones written for the model it serves. Publish
+your own with a manifest at the root of your repository. Details, the file format and the request fields:
+[docs/prompt-packs.md](docs/prompt-packs.md).
+
+## Install and run
+
+Linux, an NVIDIA GPU with 24 GB (fp8 tensor cores for the batch-invariant scheme: RTX 4090 / 5090, L4, L40S, RTX 6000
+Ada, H100), a driver supporting CUDA 13, Python 3.12.
+
+```sh
+pip install "spinf-serve[server]"
+spinf-serve serve spinf/gemma-4-12b-fp8          # downloads the checkpoint on first start; serves on port 8000
+```
+
+Then ask. Two classifications and one extraction, answered from a single read of the content:
+
+```sh
+curl http://localhost:8000/v1/score \
+  -H "Content-Type: application/json" \
+  -d '{
+    "model": "spinf/gemma-4-12b-fp8",
+    "messages": [
+      { "role": "user",
+        "content": "Customer support message:\nI was charged twice for September, please refund one of the charges." }
+    ],
+    "scoring": {
+      "queries": [
+        { "id": "topic",
+          "template": "\n\nTopic (billing, technical, account, shipping, or cancellation):{?}",
+          "options": [" billing", " technical", " account", " shipping", " cancellation"] },
+        { "id": "refund",
+          "template": "\n\nIs the customer asking for a refund?\nAnswer:{?}",
+          "options": [" yes", " no"] },
+        { "id": "month",
+          "template": "\n\nQ: Which month was charged twice?\nA: \"{*}\"" }
+      ]
+    }
+  }'
+```
+
+The response gives, for `topic` and `refund`, the probability of each answer and the same probability on empty
+content (the lift between the two is what the message says), and for `month` the most likely answers with their
+probabilities. The same body, with an API key, runs against the hosted API at `api.spinf.com`; the OpenAI SDKs work
+through `/v1/chat/completions` with the `scoring` block in `extra_body` ([quickstart](docs/quickstart.md)).
+
+As a library: `from spinf_serve import Engine; Engine("spinf/gemma-4-12b-fp8").score(request)`.
+
+## Models
+
+| name | based on | GPU | reads |
+|---|---|---|---|
+| `spinf/gemma-4-12b-fp8` (default) | Google Gemma 4 12B, FP8 with static activation scales | one 24 GB | text, images, audio, video |
+| `spinf/gemma-4-31b-fp8` | Google Gemma 4 31B, FP8 static | one 48 GB | text, images, video |
+| `spinf/qwen3.8-27b-int4` | Qwen3.8-27B, GPTQ int4 | one 24 GB | text, images, video |
+| `spinf/pplx-decider-v1-27b-int4` | Perplexity pplx-decider-v1-27b, GPTQ int4: a decision model with a 255-code readout | one 24 GB | text; scores only, with its own prompts and packs |
+
+Each is a checkpoint we quantized and calibrated, published on the Hugging Face Hub with its recipe
+(`quantizations/`); the engine pins the revision it was parity-tested on. Any other Hub checkpoint of a supported
+architecture family runs by its path, quantized at load time.
+
+## Documentation
+
+[docs/](docs/README.md): [quickstart](docs/quickstart.md), [examples](docs/examples.md) (curl calls for one question,
+parallel classifications, classification plus extraction, many contents, a prompt pack),
+[scoring API](docs/scoring-api.md), [configuration](docs/configuration.md), [writing questions](docs/writing-questions.md),
+[calibration](docs/calibration.md), [prompt packs](docs/prompt-packs.md), [FAQ](docs/faq.md); the [guides](docs/guides/index.md)
+(decision models, probabilities you can act on, few-shot classification, sentiment, survey coding, moderation, intent routing)
+and the engineering documents ([docs/README.md](docs/README.md#engineering-documents)).
+
+## Status
+
+The engine (scoring, extraction, generation, media, batch invariance, the two model stacks) is complete and
+parity-tested against recorded references. The server speaks the public API of the hosted service: `spinf-serve serve`,
+`/v1/score` and its chat-completions alias, `/v1/models`, `/v1/packs`, pack sources, the response and error envelopes.
+The `spinf/...` checkpoints are being published; [docs/README.md](docs/README.md#status) keeps the list of what is
+left. How the engine works and the measurements behind it: the engineering documents in
+[docs/](docs/README.md#engineering-documents) (architecture, batch invariance, precision, the Qwen stack, extraction).
+
+## Layout
+
+```
+spinf_serve/
+  protocol.py, compiler.py, packer.py, scorer.py, decode.py, extract.py, media.py, service.py
+                         the engine: request shape, templates to token tries, packing, one packed forward, extraction
+  manifests/             the served models (TOML): pinned checkpoint, family, schemes, capabilities, media, limits
+  packs/                 prompt packs (pure Python): the file and its validation, the evaluator, a request's packs, the sources
+                         (GitHub, folders, the official repository) and the catalog; `python -m spinf_serve.packs validate | list`
+  models/                one stack per architecture family behind one interface (base.py, registry.py): gemma4.py, qwen35/
+  quant/                 the quantization schemes (linears, kernels, the fp8-static layout), the per-revision tables, the maintainer
+                         commands that produce our checkpoints
+  server/                the HTTP server (FastAPI, one process per GPU): the public API around the engine (api.py), the settings
+                         (config.py: flags > environment > config file > manifest), the app
+  cli.py                 the `spinf-serve` command: serve, score, packs, parity, calibrate
+quantizations/           one folder per published checkpoint: recipe, calibration output, model card
+fixtures/                the public parity fixture (24 Wikinews articles x 56 cells) and one recorded reference per checkpoint
+benchmarks/              the throughput comparison script (spinf-serve against a completions server), its results, the fixture builder
+tests/                   CPU tests (run in CI); the GPU parity gate runs by hand (below)
+docs/                    the user guides and the engineering documents
+```
+
+## Develop
+
+```sh
+uv sync --extra server
+uv run pytest                                   # CPU tests (the official packs' golden fixtures need ../spinf-prompts or SPINF_PROMPTS_DIR)
+uv run ruff check . && uv run ruff format --check .
+CUDA_VISIBLE_DEVICES=0 uv run spinf-serve parity --fixture fixtures/wikinews_20261024.json.gz \
+    --reference fixtures/wikinews_20261024.spinf-gemma-4-12b-fp8.json.gz --passes 2
+```
+
+Every change that touches a kernel, a compile boundary or a quantization scheme ships with the GPU parity run:
+bit-identical log-probabilities on the recorded reference of each published checkpoint ([fixtures/](fixtures/README.md):
+24 Wikinews articles x 56 two-option cells, a reference per model recorded and verified on its card). Scores come from
+exact dependency pins; they change only together with a parity run.
+
+## Licence and acknowledgements
+
+Apache 2.0 (`LICENSE`, `NOTICE`). The models we run and re-quantize, the projects we build on, the techniques we
+implement and the data we checked on are credited in [ACKNOWLEDGEMENTS.md](ACKNOWLEDGEMENTS.md).